OBJECTIVE: To generate an algorithm to map SF-12 scores to the Euroqol EQ-5DINDEX based on responses of the U.S. Spanish-speaking Hispanics (SPHispanics).METHOD: Responses from 2386 SPHispanics who were eligible to answer the self-administered questionnaire included in the 2003 Medical Expenditure Panel Survey were used in the analyses. EQ-5DINDEX was estimated using a recently developed set of social preferences for the U.S. SPHispanics. Several ordinary least square (OLS) regression models of the EQ-5DINDEX onto the physical and mental SF-12V2 scores were developed based on 2/3 of the SPHispanics sample. Modelling was performed excluding (basic model) and including sociodemographic characteristics, and also exploring possible compound terms between SF-12V2 scores. Validation of the model was performed on 1/3 of the SPHispanics sample comparing observed and estimated EQ-5DINDEX in terms of correlation and mean absolute error (MAE).RESULTS: Based on goodness of fit, the best basic regression model considered both SF-12V2 scores plus the squared term of the mental score (R2 = 0.59). When sociodemographic characteristic were taken into account, only age significantly contributed to the previously described model (p value<0.001), but its effect on improving R2 was only marginal. Pearson correlation coefficient between observed and predicted EQ-5DINDEX was 0.77 for the best basic model in both the modelling and validating samples. MAE between observed and estimated EQ-5DINDEX was 0.076 in the validating sample.CONCLUSION: Findings confirm that reasonable estimates of the SPHispanic EQ-5DINDEX can be obtained from SF-12V2 based on a simple OLS model. This algorithm may provide researchers in Latin America with a useful method to obtained preferences-based scores when only SF-12 data is available. Given the significant amount of variance that is usually left unexplained in these types of models it would be highly recommendable to incorporate this uncertainty in sensitivity analyses.